Improving default risk prediction using Bayesian model uncertainty techniques.

摘要: Credit risk is the potential exposure of a creditor to an obligor's failure or refusal repay debt in principal interest. The measured terms probability default. Many models have been developed estimate credit risk, with rating agencies dating back 19th century. They provide their assessment default and transition probabilities various firms annual reports. Regulatory capital requirements for outlined by Basel Committee on Banking Supervision made it essential banks financial institutions develop sophisticated attempt measure higher accuracy. Bayesian framework proposed this article uses techniques physical sciences engineering dealing model uncertainty expert accuracy obtain improved estimates associated uncertainties. approach from one more incorporates historical (past performance data) estimating future probabilities. Several examples demonstrate that methodology can assess exceeding estimations all individual models. Moreover, accounts potentially significant departures “nominal predictions” due “upsetting events” such as 2008 global banking crisis.